Embedding multiple heuristics into choice models: An exploratory analysis

Contrary to the usual assumption of fixed, well-defined and context independent preferences, individuals are likely to approach a choice task using rules and heuristics that are dependent on the choice environment. More specifically, heuristics that are defined by the local choice context, such as the gains or losses of an attribute value relative to the other attributes, seem to matter significantly. Recent empirical findings also demonstrate that previous choices made by respondents and previous choice tasks shown to respondents can affect the current choice outcome, indicating a form of interdependence across choice sets. This paper reviews some of the key findings about heuristics and decision rules across the psychology, marketing, transport and environmental disciplines. Using experimental data in the context of a proposed toll road, we find that for certain components of the time and cost attributes, allowing for non-linearity and for referencing to the least desired attribute level in the local choice set offers improvement over the standard linear-in-the-attributes and linear-in-the-parameters specification. Other heuristics, including the majority of confirming dimensions and reference point revision can also be embedded into the model. An alternative approach to identifying and weighting multiple heuristics in a utility function by means of a logit-type specification for the weights is then introduced. While acknowledging the need for more work in this area, we suggest that such an approach may be a useful way of testing what happens when multiple heuristics are "mixed" in the model. (C) 2013 Elsevier Ltd. Open access under CC BY-NC-ND license.